The Effect of Withholding Antibiotics Prior to Bone Biopsy in Patients With Suspected Osteomyelitis: A Meta-analysis of the Literature

Objective. The aim of this meta-analysis is to evaluate the effect of withholding antibiotics prior to obtaining a bone biopsy in the diagnosis of osteomyelitis. Methods. Literature searches on Scopus, PubMed, and Google Scholar were performed with the input antibiotic, bone, biopsy, osteomyelitis, and prior both separately and together, though initially as a combination, as key words for all study types in the English language published up until July 2018. The initial combination search (and subsequent additional search of the references) yielded 924 related articles, which ultimately resulted in 12 articles with adequate data for analysis. Data were analyzed using an inverse variance method to determine the weight of the studies in a random effects model, pooling the data for odds ratios (OR) and heterogeneity. Bias risk was determined with the Quality Assessment of Diagnostic Accuracy Studies 2 system. Results. Of the 12 studies, 8 failed to demonstrate that antibiotics administered prior to bone biopsy have an effect on culture yield, while 4 reported a significant effect on culture yield. The total weighted OR for all studies was 2.02 (95% confidence interval [CI], 0.94–4.36; P = .04) and the prediction interval was 0.17–23.63. When vertebral osteomyelitis (VO) and nonvertebral osteomyelitis (NVO) articles were evaluated separately, the VO OR was 2.95 (95% CI, 1.40–6.24) and the prediction interval was 0.40–21.53. The OR for NVO was 0.66 (95% CI, 0.04–12.03) and the prediction interval was 0–114.53. Conclusions. When all studies as well as when NVO studies only were evaluated, there was no significant difference in bacterial pathogen yield based on antibiotic exposure. When VO studies were evaluated separately, there were fewer pathogens recovered than when antibiotics were administered prior to obtaining bone cultures.

RELATED CONTENT

Do Antibiotic-impregnated Calcium Sulfate Beads Improve the Healing of Neuropathic Foot Ulcers With Osteomyelitis Undergoing Surgical Debridement?

Diagnosing Diabetic Foot Osteomyelitis

Comparison Between Tc-99m WBC SPECT/CT and MRI for the Diagnosis of Biopsy-proven Diabetic Foot Osteomyelitis

There is a strong belief that antibiotics should be held prior to bone culture. This theory stems from the thought that prebiopsy antibiotics will penetrate into the bone and interfere with bacterial pathogen recovery and identification with traditional cultures; however, the results of current publications are inconsistent regarding whether this theory should be supported or rejected.^1-12 If antibiotics are unnecessarily withheld, especially in cases of adjacent abscess, compromised host, or open wound, the delay could result in sepsis, bacteremia, and worse clinical outcomes.

Most of the published work in this area are from patients with hematogenous vertebral osteomyelitis (VO). The theory that holding antibiotics is necessary to obtain “true” culture results has been generalized to other types of contiguous osteomyelitis such as diabetic foot, pressure ulcer, and trauma.¹³ While the disease is similar in that it is an infection of the bone, there are several differences in the overall disease pathology: etiology (hematogenous spread vs. contiguous spread), patient demographics, risk factors, comorbidities, time to presentation, concomitant soft tissue necrosis and abscess, clinical outcomes, and treatment protocols. With multiple types of osteomyelitis, it begs the question if one type of osteomyelitis can be generalizable to other types.

Several practice guidelines^13-17 recommend stopping antibiotics before bone cultures are obtained but recognize the strength of the literature to support this recommendation is low or does not comment on holding antibiotics.¹⁸ Some recommend withholding antibiotics, even when the results of the study were negative.^7,9,19 Others note the time period recommended to withhold antibiotics is arbitrary⁶ and the time recommended to hold antibiotics ranges from several days to weeks.^13-15,20 In contrast, the orthopedic literature endorses administration of empirical antibiotics in trauma situations with the understanding that intraoperative biopsy of the bone will be used to target the antibiotics to the infecting pathogens.^21-23 The present authors have been unable to identify any literature that has evidence to support withholding antibiotics in diabetic foot infections, trauma, pressure ulcers, or other causes of direct extension osteomyelitis. The aim of this study is to evaluate the current literature to determine if sufficient evidence is present to recommend withholding prebiopsy antibiotics as a step to increase traditional culture success.

This was a meta-analysis of the current, publicly available literature and did not require formal ethical approval. No patient information was encountered or handled.

Literature searches on Scopus, PubMed, and Google Scholar were performed with the input antibiotic, bone, biopsy, osteomyelitis, and prior separately and together, though initially as a combination, as key words for all study types in the English language published up until July 2018. The combined, and later additional, search initially yielded 924 articles; these articles were evaluated by 2 reviewers (P.A.C. and L.A.L.). Thirteen papers were identified that compared antibiotic exposure with bacterial culture yield and included both a group being exposed to antibiotics before bone biopsy and a group not being exposed to antibiotics prior to bone biopsy. One paper²⁴ did not have adequate data to evaluate the stated outcomes, resulting in 12 articles for evaluation (Figure 1). References were evaluated for related articles and were included when appropriate. 

The risk of bias was evaluated with the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2; University of Bristol, Bristol, United Kingdom) scoring system by 2 investigators (P.A.C. and L.A.L.).²⁵ Of the 12 papers, 10 had a high risk of bias^1-10 and 2 had a low risk.^11,12 There was full inter-reviewer agreement for 50 of 52 of the subsections, with a κ = 0.92 (95% confidence interval [CI], 0.82–1.00) indicating a very good agreement. Final assignment of the risk of bias was determined with the modified Delphi method (Table 1^1-12). Each analysis had 4 subgroups that helped guide the overall rating of each article as low, high, or unknown risk of bias. Any discrepancies were addressed with a modified Delphi method and a consensus was obtained between the 2 reviewers (P.A.C. and L.A.L.).²⁶ If consensus was not obtained, further input was obtained from a third reviewer (J.L.). Inter-rater agreement was determined using the kappa test.

The data from the articles were entered into Table 1,^1-12 including the reference standard used to diagnose osteomyelitis. There was considerable variation in the reference standard used to diagnosis osteomyelitis, and no 2 studies used the same reference standard. Reference standards varied widely and included, but were not limited to, magnetic resonance imaging; computed tomography; evaluation of symptoms or histological, radiological, or pathological features; and blood or bone culture (Table 1^1-12).

Two authors (P.A.C. and L.A.L.) repeated the chi-square analysis performed in each paper to compare bacterial yield in patients who received antibiotics prior to biopsy and patients who did not. If discrepancies were found, they were noted in Table 1.^1-12 

A pooled, weighted analysis (χ²) was performed of the data using the Meta-Essentials Excel package.^27,28 These data were analyzed in 3 ways. First, data from the 12 studies were combined using an inverse variance method to determine the appropriate weight of the studies in a random effects model, as the studies draw from populations that differ from each other (varied etiology and anatomic location) in ways that could affect the outcome²⁹ (Table 2^1-12). This was repeated for the VO studies (n = 9; Table 3^1-9) and for the nonvertebral osteomyelitis (NVO) studies (n = 3; Table 4^10-12). Effect size measure was evaluated along with odds ratios (OR) for each of these groups (Table 2, Table 3, and Table 4^1-12). The effect size was plotted on a logarithmic scale Forest plot with 95% CI (Figure 2, Figure 3, and Figure 4^1-12). Cochran Q and P_Q were used to determine the presence of heterogeneity, and the magnitude of heterogeneity was determined by I².^27,28,30 In addition,τ² and τ were calculated, where τ estimates standard deviation of the distribution of true effect sizes.^31,32

Twelve articles were identified from a review of the available literature that retained adequate data for comparison.^1-12 Four studies reported antecedent antibiotic treatment was associated with reduced bacterial cultures,^3-5,8 and 8 failed to identify a significant relationship between antibiotic exposure prior to bone biopsy and the success of culture^1,2,6,7,9-12 (Table 1^1-12). All studies were retrospective except for 1 prospective study¹² of patients with diabetic foot infections. 

Statistics presented in the studies were independently evaluated by 2 investigators (P.A.C. and L.A.L.). Incorrect results were identified in 2 papers.^7,9 Rankin et al⁷ did not provide any statistical analysis beyond simple percentages, and, although significance was not claimed, they endorsed that antibiotics prior to biopsy were likely to reduce culture yield. They⁷ reported 2 of 8 patients (25%) who received antibiotics had positive cultures, while 6 of 12 patients (50%) not receiving antibiotics had positive cultures. When their data⁷ were evaluated with Fischer’s exact test, (given the small sample size) the results were not significant (25% vs. 50%; P = .37). Wang et al⁹ reported a significantly lower yield of cultures in patients who received antibiotics prior to biopsy (11/17, 64.7%) compared with patients who did not receive prebiopsy antibiotics (21/24, 87.5%). When independently reevaluated, their reported P value of .04 was found to be P = .13.⁹ These changes resulted in 2 articles^7,9 that no longer support this theory of prebiopsy antibiotics reducing the success of bacterial cultures.

Data were analyzed using an inverse variance method to determine the appropriate weight of the studies in a random effects model. The Forest plots (Figure 2, Figure 3, and Figure 4^1-12) depict a graphical representation of the relation of the sample size, CI, and relative study weight as represented by the area of the circles in the figures. The vertical line is the line of no effect and indicates no significant effect when the CI crosses it. Line 13 in Figure 2,^1-12 line 10 in Figure 3,^1-9 and line 4 in Figure 4^10-12 are the weighted combined effects of the 12 (Table 2^1-12), 9 (Table 3^1-9), and 3 (Table 4^10-12) studies, respectively. The OR for withholding antibiotics in all studies was 2.02 (95% CI, 0.94–4.36) and the prediction interval was 0.17 to 23.63. Both CIs crossed the line of no effect. When the VO and NVO papers were evaluated separately, the VO OR was 2.95 (95% CI, 1.40–6.24) and the prediction interval was 0.40 to 21.53 (Figure 3^1-9). The NVO OR was 0.66 (95% CI, 0.04–12.03) and the prediction interval was 0 to 114.53 (Figure 4^10-12). 

Study heterogeneity was evaluated using Cochran’s Q, I², τ², and τ. For all studies, a Q value of 44.99 was determined with a P_Q< .0001. This indicated that heterogeneity exists among the studies. The I² (proportion of observed variance) was found to be 75.6%, which suggested the populations are not similar.³⁰ The τ² and τ were 1.13 and 1.06, respectively. Given the large amount of heterogeneity, evaluation of publication bias could not be completed properly.³² Given the heterogeneity and results of the Forest plot, the combined data failed to identify a significant effect of antibiotics on the culture yield for bone biopsy to diagnose osteomyelitis. 

Further subgroup analysis was performed to isolate the VO studies (Table 3^1-9). For the VO studies, a Q value of 20.7 was found with a P_Q= .008 and I² of 61.4%, indicating heterogeneity and dissimilar populations. The τ² and τ were 0.64 and 0.80, respectively. The NVO analysis (Table 4^10-12) also exhibited heterogeneity (Q = 8.4, P_Q= .015, I² = 76.0%, τ² = 0.98, τ = 0.99).

This is the first meta-analysis to evaluate this common clinical question. The results of the meta-analysis suggest there was a significant difference in culture yield in patients with VO that received antibiotics prior to bone biopsy compared with patients without antibiotics. However, there was significant heterogeneity among the studies, and the prediction interval was large in each analysis (Figure 2, Figure 3, and Figure 4^1-12). There was no difference in NVO studies (Figure 4^10-12).

Many physicians have strong feelings about stopping antibiotics before cultures are obtained as well as the impact of any antibiotic treatment of the reliability of cultures. Most of this belief is based on small studies of patients with hematogenous VO that have been generalized to osteomyelitis at other sites and other mechanisms of infection. It would not be acceptable to make this generalization, just as it would be unacceptable to combine all of the studies from different anatomic sites into 1 meta-analysis, as demonstrated in Table 1.^1-12

There was considerable variability and bias in the published studies based on the reference standard to define osteomyelitis, sample size, patient characteristics, and statistical analysis of the data. One of the most often quoted papers is a letter to the editor, which ultimately failed to identify significance that antibiotics prior to biopsy reduce culture recovery.¹⁹ It was not included in the current evaluation of studies^1-12 because it did not have adequate information about the study methods. 

When the literature recommends withholding antibiotics, the duration of time to hold the antibiotics is inconsistent. In this analysis, no definitive temporal association of antibiotic exposure and the bone biopsy was made in 5 studies.^2,7,9,11,12 One study classified having antibiotic exposure by occurring within 24 hours of the biopsy,¹⁰ 1 within 48 hours of the biopsy,³ and 2 reported a median exposure to antibiotics within 4 days of the biopsy with a range of exposure of 1 to 20 days⁵ and 1 to 37 days.⁶ One study classified a patient as being exposed to antibiotics if they had antibiotic administration within 6 weeks of the biopsy¹ and 1 within 6 months prior to biopsy.⁸

The majority of the literature addressing the effect of prebiopsy antibiotic exposure focuses on VO,^1-9 with only 3 studies^10-12 that evaluate NVO. The VO literature^1-9 supports the commonly held belief that prebiopsy antibiotics negatively impact the reliability of culture yield with bone biopsy. In fact, only VO studies support this theory. Several VO studies^1,2,6,7,9 and all of the NVO studies^10-12 were negative. When the NVO studies are evaluated separately, this commonly held belief is not supported. 

The result of hematogenous osteomyelitis studies are not generalizable to other types of osteomyelitis. For instance, diabetic foot osteomyelitis is often seen in an older patient population that is predominately male with severe sensory neuropathy, immunopathy, micro- and macrovascular diseases, poor glucose control, chronic kidney disease, and malnutrition. Diabetic foot osteomyelitis is almost always preceded by a penetrating ulceration and is sometimes accompanied by abscess or soft tissue necrosis.^13,15 In contrast, VO is often identified in younger patients with few comorbidities, no wounds, and no vascular impairment involving the site of infection.^14,33 The average culture yield in patients with VO was 34.2% in patients exposed to antibiotics and 53.6% in patients without antibiotic exposure, while patients with diabetic foot osteomyelitis who were exposed to antibiotics had a reported positive culture rate of 87.8% and 61.8% for patients without exposure (Table 1^1-12). In addition, studies report widely varying pathogen recovery ranging from 29% to 91% in patients with VO^34-36 and 23.5% to 87.6% in other etiologies.^10-12

Misclassification of osteomyelitis is a primary underlying concern because the reference standards were poor, ill-defined, or simply not provided at all in the methods of the paper¹⁴ (Table 1^1-12). Many studies used multiple criteria as the reference standard, and none of these studies provided specific information about how many cases were selected based on the different criteria. None of the studies used bone histology alone, a recognized gold standard, as the reference standard.^13,15 Poor reference standards increased the risk of misclassifying osteomyelitis. In addition, the bacterial yield was very low in several studies, even in patients who had not received antibiotics; the low yield of bacterial cultures may simply be a result of including patients that did not have osteomyelitis rather than as a result of antibiotics changing culture results. 

A prospective study by Cecilia-Matilla et al¹² in the diabetic foot osteomyelitis literature observed that antibiotics (mean duration of exposure, 6 ± 4.1 weeks) did not have a negative impact on microbiologic culture results and actually showed an increased yield in patients who were exposed to antibiotics as opposed to those who were antibiotic naïve (87.8% vs. 61.8%, respectively; P < .001). This phenomenon also was exhibited by Marschall et al⁶ and Zhorne et al¹¹ in patients with VO and VO and NVO, respectively, who reported higher rates of pathogen identification and treated with antibiotics before bone cultures were obtained compared with patients without antibiotic exposure (72.0% vs. 53.1%, P = .14; and 70.0% vs. 63%, P = .55, respectively). 

The method of biopsy uniformly has been shown to impact pathogen recovery. Open surgical bone biopsy provides higher yields of bacterial pathogens than percutaneous bone biopsy. While this may be related to superior biopsy specimen methods, it also may be due to the higher pretest probability of having osteomyelitis given the surgeon’s decision to proceed with surgical intervention. Most of the studies used percutaneous needle biopsy^1-4,6-10,24 to obtain specimens. Unfortunately, some studies^6,14,37 do not delineate the proportion of cultures that were obtained with opened and percutaneous biopsy and the yield with each approach. Marschall et al⁶ delineated between percutaneous and open biopsies and reported a significant increase in culture yield in patients who had open surgical biopsy compared with patients who had percutaneous biopsy, including those who were exposed to antibiotics prior to the biopsy (95% vs. 59%, respectively; P = .02) and patients who were not (82% vs. 43%, respectively; P = .04). Saravolatz et al³⁷ evaluated differences in culture yield between open and percutaneous biopsies and reported a higher yield in open biopsies, but no significance difference was identified between culture yield in patients that had pre-antibiotic exposure and no antibiotic exposure. 

Many clinicians have a strong belief that antibiotic exposure makes bone culture results unreliable, even though the majority of the published work does not support this position. The practice guidelines for VO, diabetic foot osteomyelitis, and bone and joint infections endorse withholding antibiotics prior to bone biopsy, but the studies referenced to support this recommendation do not actually support the position.^13-17,24 The results of this meta-analysis failed to reject the null hypothesis that prebiopsy administration of antibiotics change the bacterial yield of traditional cultures when the literature was evaluated as a whole. Clearly, there is a need for prospective studies to rigorously evaluate this important question in specific anatomic sites and mechanisms of infection in patients suspected of osteomyelitis.

Note: Appreciation is given to Yin Xi, PhD, for his contribution to the statistical analysis of this manuscript.

Authors: Peter Andrew Crisologo, DPM¹; Javier La Fontaine, DPM, MS¹; Dane K. Wukich, MD²; Paul Kim, DPM, MS^1,3; Orhan K. Oz, MD, PhD⁴; and Lawrence A. Lavery, DPM, MPH¹

Contributions: Dr. Crisologo participated in the literature search, article review, statistical analysis, and writing of the manuscript. Dr. La Fontaine participated in the statistical analysis, writing, and critical review of the manuscript. Dr. Kim participated in the writing and critical review of the manuscript. Dr. Wukich participated in the writing and critical review of the manuscript. Dr. Oz participated in article review, statistical analysis, writing, and critical review of the manuscript. Dr. Lavery participated in the literature search, article review, statistical analysis, and writing of the manuscript. 

Affiliations: ¹Department of Plastic Surgery, University of Texas Southwestern Medical Center, Dallas, TX; ²Department of Orthopaedic Surgery, University of Texas Southwestern Medical Center; ³Department of Plastic Surgery, MedStar Georgetown University Hospital, Washington, DC; and ⁴Department of Radiology, University of Texas Southwestern Medical Center

Correspondence: Peter A. Crisologo, DPM, Department of Plastic Surgery, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX, 75390; Andrew.crisologo@gmail.com 

Disclosure: This work was supported by the National Institutes of Health (3 U24 DK076169-08S4) and the American Diabetes Association (1-17-ICTS-056). The authors do not have any relevant conflicts of interest.